huggingface-lora-space-builder

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly instructs the agent to ask the user to "Paste it here" for a Hugging Face access token and includes examples that use that token value (e.g., passing it to HfApi and add_space_secret), which requires the LLM to receive and potentially echo or embed the secret in outputs — an insecure pattern for secret handling.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The workflow ingests outsider-authored free text from Hugging Face LoRA model repos by loading README.md/model-card content at runtime (e.g., ModelCard.load(repo_id) and card.text / README body) to extract trigger words, hyperparameters, and example prompts/media.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs adding/installing remote Git packages at build/runtime (e.g. "git+https://github.com/huggingface/diffusers" and "git+https://github.com/Lightricks/LTX-Video.git"), which will fetch and execute remote code as required dependencies during Space build/runtime.